
On 2006-12-09 16:19, Peter Boncz wrote:
Update of /cvsroot/monetdb/pathfinder/runtime In directory sc8-pr-cvs7.sourceforge.net:/tmp/cvs-serv28149
Modified Files: Tag: XQuery_0-14 pathfinder.mx pf_support.mx Log Message: various performance fixes; small inserts and deletes are now reasonably fast
- index new qnames in qnloc index - better handling of the hash tables on the qn_loc index - shortened the collection lock time by moving most of the logging out of the critical period
Index: pf_support.mx =================================================================== RCS file: /cvsroot/monetdb/pathfinder/runtime/pf_support.mx,v retrieving revision 1.172.2.11 retrieving revision 1.172.2.12 diff -u -d -r1.172.2.11 -r1.172.2.12 --- pf_support.mx 8 Dec 2006 17:48:43 -0000 1.172.2.11 +++ pf_support.mx 9 Dec 2006 15:19:52 -0000 1.172.2.12
@@ -2096,47 +2097,22 @@ }
var texts := ws.fetch(UPDATED_TEXT); - fix_consecutive_texts(ws, texts); + #fix_consecutive_texts(ws, texts);
# get containers in the order in which we must commit them ([COLL_ID,CONT]) var cont_order := affected_conts.reverse().mirror().leftfetchjoin(ws.fetch(CONT_COLL)).reverse().sort();
1. what does this change have to do with the log message? 2. this change breaks the document: if a node between two text nodes is deleted, the two text nodes have to be coalesced. That is what this function does. It is a necessary step to keep subsequent queries working correctly. -- Sjoerd Mullender